WINE DATASET EXPLORATORY DATA ANALYSIS

  • TERM 1 and TERM 2 EDA PROJECT (INSAID ACADEMY)
  • Submitted by Jatinder Mahajan (Aug,10 2019 Batch)

Table of Contents

  1. Problem Statement
  2. Importing Packages
  3. Loading the Dataset
  4. Understanding the Data
  5. Data Processing
  6. Exploratory Data Analysis
  7. Conclusion
  8. Actionable Insights

1. Problem Statement :

Wine is globally popular Alcoholic beverage which is made from different varieties of grapes (red & white) by fermentation of sugar(grape juice) to ethanol,carbon dioxide (CO2) and heat. Different types of grapes and varied strains of Yeast produce different variations of wines. These variations result from the complex interactions between the biochemical development of the grape, the reactions involved in fermentation, the terroir(environmental characteristics of crop), and the production process.

We will be doing Exploratory Data Analysis of the available wine datasets to derive insights on how different characteristics derive the quality and taste of the wine. How these features are interlinked and how the quality varies with different levels of these ingredients.

2. Import Packages

In [2]:
import numpy as np                                 # Numpy is Python Package to work with numerical Data

import pandas as pd                                # Pandas Package will assist in working with Dataframes
pd.set_option('mode.chained_assignment', None)      # To suppress pandas warnings.
pd.set_option('display.max_colwidth', -1)           # To display all the data in each column
pd.options.display.max_columns = 50                 # To display every column of the dataset in head()

import warnings
warnings.filterwarnings('ignore')                   # To suppress all the warnings in the notebook.


import matplotlib.pyplot as plt
%matplotlib inline

import seaborn as sns
sns.set(style='whitegrid', font_scale=1.3, color_codes=True)      # To apply seaborn styles to the plots.

3. Loading the Dataset

Dataset is provided by INSAID and taken from their github repository. Reference link : https://github.com/insaid2018/Term-1/blob/master/Data/Projects/winequality.csv

In [3]:
df_wine=pd.read_csv('https://raw.githubusercontent.com/insaid2018/Term-1/master/Data/Projects/winequality.csv')

4. Understanding the Data

Let us understand our data, how it is distributed and what are the different variables it has.

In [4]:
df_wine.head(5)  ### Displaying the First Five Rows of the data.
Out[4]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality
0 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5
1 7.8 0.88 0.00 2.6 0.098 25.0 67.0 0.9968 3.20 0.68 9.8 5
2 7.8 0.76 0.04 2.3 0.092 15.0 54.0 0.9970 3.26 0.65 9.8 5
3 11.2 0.28 0.56 1.9 0.075 17.0 60.0 0.9980 3.16 0.58 9.8 6
4 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5

The dataset input variables (based on physicochemical tests) are:

  1. fixed acidity (tartaric acid - g / dm^3)
  2. volatile acidity (acetic acid - g / dm^3)
  3. citric acid (g / dm^3)
  4. residual sugar (g / dm^3)
  5. chlorides (sodium chloride - g / dm^3
  6. free sulfur dioxide (mg / dm^3)
  7. total sulfur dioxide (mg / dm^3)
  8. density (g / cm^3)
  9. pH
  10. sulphates (potassium sulphate - g / dm^3)
  11. alcohol (% by volume)
  12. quality (score between 0 and 10)
  • Taken this units from web research
In [5]:
print("Shape of the wine dataset {s}".format(s=df_wine.shape))  ## Printing the shape of the dataset

Shape of the wine dataset (6497, 12)
From this we know that our dataset has 12 variables and 6497 values.

4.1 Pre Profiling Analysis

In [6]:
df_wine.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 6497 entries, 0 to 6496
Data columns (total 12 columns):
fixed acidity           6497 non-null float64
volatile acidity        6497 non-null float64
citric acid             6497 non-null float64
residual sugar          6497 non-null float64
chlorides               6497 non-null float64
free sulfur dioxide     6497 non-null float64
total sulfur dioxide    6497 non-null float64
density                 6497 non-null float64
pH                      6497 non-null float64
sulphates               6497 non-null float64
alcohol                 6497 non-null float64
quality                 6497 non-null int64
dtypes: float64(11), int64(1)
memory usage: 609.2 KB

This has given us the preview of our data which is:

  • 12 different variables(columns) with 6497 observations(rows),
  • all columns are with numeric datatype.
In [7]:
df_wine.describe()
Out[7]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality
count 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000 6497.000000
mean 7.215307 0.339666 0.318633 5.443235 0.056034 30.525319 115.744574 0.994697 3.218501 0.531268 10.491801 5.818378
std 1.296434 0.164636 0.145318 4.757804 0.035034 17.749400 56.521855 0.002999 0.160787 0.148806 1.192712 0.873255
min 3.800000 0.080000 0.000000 0.600000 0.009000 1.000000 6.000000 0.987110 2.720000 0.220000 8.000000 3.000000
25% 6.400000 0.230000 0.250000 1.800000 0.038000 17.000000 77.000000 0.992340 3.110000 0.430000 9.500000 5.000000
50% 7.000000 0.290000 0.310000 3.000000 0.047000 29.000000 118.000000 0.994890 3.210000 0.510000 10.300000 6.000000
75% 7.700000 0.400000 0.390000 8.100000 0.065000 41.000000 156.000000 0.996990 3.320000 0.600000 11.300000 6.000000
max 15.900000 1.580000 1.660000 65.800000 0.611000 289.000000 440.000000 1.038980 4.010000 2.000000 14.900000 9.000000
This is has provided us the descriptive statistiscal information about our dataset.
It also gives information that the dataset is very clean without any abnormal, missing values and negative values.
  • Most variables have mean values more than median which is 50Th percentile.
  • We also observe that there is considerable difference in values of 75th Percentile and max value for these columns “residual sugar”,”free sulfur dioxide”,”total sulfur dioxide”.
  • Above observations suggests that there are Outliers in our data set.

4.1.1 Confirming Missing Values

Although, we have seen above that we don't have a missing value, however we can confrim if we have any null value in any of the variable.

In [8]:
df_wine.isnull().sum()
Out[8]:
fixed acidity           0
volatile acidity        0
citric acid             0
residual sugar          0
chlorides               0
free sulfur dioxide     0
total sulfur dioxide    0
density                 0
pH                      0
sulphates               0
alcohol                 0
quality                 0
dtype: int64

This confirms there are no missing values in our dataset.

4.2 Profiling Analysis

Pandas Profiling gives illustrative picture of the dataset. It helps to identify the variables to derive insights.

In [9]:
import pandas_profiling        # import pandas_profiling package
In [143]:
profile=df_wine.profile_report(title='Pandas Profiling of Wine Dataset')
profile.to_file(output_file="profiling_wine_dataset.html") # Storing the Profile in separate HTML File 

df_wine.profile_report(title='Pandas Profiling of Wine Dataset', style={'full_width':True})  # To display in notebook.
Out[143]:





















Profiling of the dataset has shown us that dataset is Clean without any abnormal values, missing values or negative values.
It also gave the slight idea that we have 11 predictive variables and last one 'quality' a response variable .


During the course of our data explorataion and analysis , we will see how quality is getting affected from other variables which is also our problem statement.



















5. Data Processing
Let's rename our columns to get rid of spaces in their name. It will be easy later to play with the dataset.













In [11]:



    


df_wine.rename(columns={'fixed acidity': 'fixed_acidity','citric acid':'citric_acid','volatile acidity':'volatile_acidity','residual sugar':'residual_sugar','free sulfur dioxide':'free_sulfur_dioxide','total sulfur dioxide':'total_sulfur_dioxide'}, inplace=True)
df_wine.head(n=5)    ### Display dataset with updated column names.



    















    
Out[11]:











  
    

      
      fixed_acidity
      volatile_acidity
      citric_acid
      residual_sugar
      chlorides
      free_sulfur_dioxide
      total_sulfur_dioxide
      density
      pH
      sulphates
      alcohol
      quality
    

  
  
    

      0
      7.4
      0.70
      0.00
      1.9
      0.076
      11.0
      34.0
      0.9978
      3.51
      0.56
      9.4
      5
    

    

      1
      7.8
      0.88
      0.00
      2.6
      0.098
      25.0
      67.0
      0.9968
      3.20
      0.68
      9.8
      5
    

    

      2
      7.8
      0.76
      0.04
      2.3
      0.092
      15.0
      54.0
      0.9970
      3.26
      0.65
      9.8
      5
    

    

      3
      11.2
      0.28
      0.56
      1.9
      0.075
      17.0
      60.0
      0.9980
      3.16
      0.58
      9.8
      6
    

    

      4
      7.4
      0.70
      0.00
      1.9
      0.076
      11.0
      34.0
      0.9978
      3.51
      0.56
      9.4
      5
    

  









































6. Exploratory DATA ANALYSIS



Lets dig deep to derive insights from our dataset.
































6.1 Univariate Analysis



Univariate analysis or visualization will give the insights about the data distribution of the each variable.
 It will provide statistics of each individual variable.

















6.1.1 Histograms
Lets plot histograms for each variable to understand the underlying distribution of













In [12]:



    


df_wine.hist(figsize=(20,15),color='skyblue')
plt.show()



    















    






























6.1.2 Density plot to check linearity
To check the linearity of the variables it is a good practice to plot distribution graph and look for skewness of features.
Kernel density estimate (kde) is a quite useful tool for plotting the shape of a distribution













In [13]:



    


sns.set(rc={'figure.figsize':(5,4)})
for i in df_wine.columns:
    sns.distplot(df_wine[i])
    plt.title(i)
    plt.show()



    















    

















    

















    

















    

















    

















    

















    

















    

















    

















    

















    

















    




























    

  • 'pH' column appears to be normally distributed
    

    • 

  • Rest all  independent variables are right skewed/positively skewed
    

    • 



The above distribution analysis shows the main feature is quality, against which all other features can be measured.
Other important features could be alcohol, acidity, sulfur dioxide and pH .

















Let us explore more on 'quality'
Check for the unique values of the quality













In [14]:



    


df_wine['quality'].unique()      #### Checking unique values withing quality variable.



    















    
Out[14]:








array([5, 6, 7, 4, 8, 3, 9], dtype=int64)



















In [15]:



    


'''
 Checking how data is distibuted among these unique quality values
 and arranging them in ascending order
 
'''

df_wine.quality.value_counts().sort_index()  



    















    
Out[15]:








3    30  
4    216 
5    2138
6    2836
7    1079
8    193 
9    5   
Name: quality, dtype: int64



















In [16]:



    


fig = plt.figure(figsize = (10,6))
plt.hist(df_wine["quality"].values, range=(1, 10),color='darkred')
plt.xlabel('Ratings of wines')
plt.ylabel('Count')
plt.title('Distribution of wine ratings')
plt.show()



    















    




























We can see quality has a range from 3-9 with maximum observations falls between 5 to 6 and rare observation with high quality from 8-9.



















6.2 Bivariate Analysis


















6.2.1 Categorizing Quality Variable.
Since crux of our dataset lies on Quality , let's distribute quality in to categorical variable called "Rating"
Based on Quality Values we can distribute as following :


    

  • bad : Quality values less than 4
    • 

  • Average : Quality values between 4 to 7
    • 

  • Good : Quality values above 7
    • 














In [111]:



    


'''
Adding a new column'rating' to our dataset 
Bifurcating Quality column based on values mentioned above. 
Defining good and bad and keep all others in average category as default

'''

df_wine['rating']=np.select([(df_wine['quality'] <= 4),(df_wine['quality'] >= 7)],['bad','good'],default='average')

# Displaying data distribution in 'rating' column

df_wine['rating'].value_counts().plot(kind='pie',autopct='%5.1f%%',wedgeprops=dict(width=0.15),  
                                                                  figsize=(10, 8), fontsize=11, 
                                                                 startangle=20, shadow=True, cmap='binary')
plt.title('Rating Distribution',fontsize=16)



    















    
Out[111]:








Text(0.5, 1.0, 'Rating Distribution')










    




























Above Donut Chart shows us majority of the data lies under "Average" Categrory.

















6.2.2 Correlation Among Variables
Let's find Correlation among the variables













In [145]:



    


mycor=df_wine.corr()
plt.subplots(figsize=(10,10))
sns.heatmap(mycor,annot=True,cmap='RdPu_r',vmax=1)
plt.title('Heat Map showing different Correaltion Among Variables',fontsize=16)



    















    
Out[145]:








Text(0.5, 1.0, 'Heat Map showing different Correaltion Among Variables')










    




























    

  • We can observe that, the 'alcohol, sulphates, citric_acid & fixed_acidity' have maximum corelation with response variable 'quality'.
    • 

  • Also, we can 'fixed_acidity,volatile_acidity,residual_sugar,chlorides,density have negative corelation with the 'quality'.
    • 

  • We also observe that alcohol has very strong negative corelation with density.
    • 

  • pH vs fixed acidity(negative correlation)
    • 

  • density vs fixed acidity (positive correlation)
    • 

  • citric acid vs fixed acidity(positive correlation)
    • 

  • citric acid vs volatile acidity(negative correlation)
    • 

  • free sulphur dioxide vs total sulphur dioxide(positive correlation)
    • 

  • density vs alcohol(negative correlation)
    • 

  • This means that, they need to be further analysed for detailed pattern and corelation exploration. 
    • 














In [41]:



    


# Let's check the correlation values of quality with other variables.

mycor['quality'].sort_values(ascending=False) 



    















    
Out[41]:








quality                 1.000000
alcohol                 0.444319
citric_acid             0.085532
free_sulfur_dioxide     0.055463
sulphates               0.038485
pH                      0.019506
residual_sugar         -0.036980
total_sulfur_dioxide   -0.041385
fixed_acidity          -0.076743
chlorides              -0.200666
volatile_acidity       -0.265699
density                -0.305858
Name: quality, dtype: float64






































6.2.3 Checking Correlation between different Variables using different Plot styles












In [106]:



    


sns.jointplot(x=df_wine["quality"],y=df_wine["pH"],kind='scatter',color='red')
plt.title('Variation of Quality with PH level',loc='left')



    















    
Out[106]:








Text(0.0, 1.0, 'Variation of Quality with PH level')










    




























As we observe from the above analysis, there are the majority of wines falling into pH window of between 3.3 and 3.6, but it is not just the good quality wines. This suggests that this is a common basic feature of wine, rather than an indicator of good red wine.













In [89]:



    


bx = sns.factorplot('quality','alcohol',data=df_wine,color='red')
bx.set(xlabel='Wine Ratings', ylabel='Alcohol Percentage', title='Alcohol percentage (ABV)in different types of Wine ratings')



    















    
Out[89]:








<seaborn.axisgrid.FacetGrid at 0x198255c9400>










    




























We observed with above graph that , good quality wines have around 11.5-12.5 % of Alcohol and bad quality wines have lower alcohol with median value around 10.













In [74]:



    


sns.jointplot(x=df_wine["quality"],y=df_wine["total_sulfur_dioxide"],kind='scatter',color='red')



    















    
Out[74]:








<seaborn.axisgrid.JointGrid at 0x1980c802828>










    




























SO2 or Sulphur Dioxide is widely used in wine making process as it lower the oxidation and kills the harmful bacteria, however having more sulphites in wine does not make it good in quality. Average wines have 250- 350 PPM of sulphur dioxide, bad quality wines have higher percentage , some above 400 as well and good wines have lower sulphur dioxide percentage as low as 90. So it is evident that use of Sulphur Dioxide needs to moderated.















Lets analyse some negative corelations













In [103]:



    


sns.jointplot(x=df_wine["quality"],y=df_wine["residual_sugar"],kind='scatter',color='red')
plt.title('Quality Vs Residual Sugar',loc='left')



    















    
Out[103]:








Text(0.0, 1.0, 'Quality Vs Residual Sugar')










    




























residual_sugar is what is left of sugar after interacting with yeast in fermentation process. The above graph shows lesser the residual_sugar better is the quality of wine. Even most of the average wines have less residual_sugar.But as we see the bad wines also have less residual_sugar , so we cannot conclude on this as the most promising factor of having good wine, although its influence is considerable.













In [83]:



    


sns.boxplot(x='quality',y='fixed_acidity',data=df_wine,fliersize=5)



    















    
Out[83]:








<matplotlib.axes._subplots.AxesSubplot at 0x19824eb4d68>










    




























As mentioned above fixed_acidity has a negative correlation of with quality. Most of the bad quality wines have higher fixed_acidity values as compare to good quality wine.













In [101]:



    


sns.jointplot(x=df_wine["density"],y=df_wine["alcohol"],kind='kde',color='red')
plt.title('Density Plot showing variation of Alcohol with Density of Wine',loc='left')



    















    
Out[101]:








Text(0.0, 1.0, 'Density Plot showing variation of Alcohol with Density of Wine')










    




























As we see, alcohol has a strong negative corelation with density, higher the percentage of alcohol have lower density.













In [92]:



    


bx = sns.violinplot(x="rating", y='citric_acid', data = df_wine)
bx.set(xlabel='Wine Ratings', ylabel='Citric Acid', title='Citric_Acid in different types of Wine ratings')



    















    
Out[92]:








[Text(0, 0.5, 'Citric Acid'),
 Text(0.5, 0, 'Wine Ratings'),
 Text(0.5, 1.0, 'Citric_Acid in different types of Wine ratings')]










    


























In [95]:



    


sns.regplot(data=df_wine, x='quality', y='volatile_acidity', color='brown')
plt.title('Regplot showing the variation of quality with volatile Acidity')



    















    
Out[95]:








Text(0.5, 1.0, 'Regplot showing the variation of quality with volatile Acidity')










    




























Volatile acidity (VA) is a measure of the wine's volatile (or gaseous) acids. The primary volatile acid in wine is acetic acid, which is also the primary acid associated with the smell and taste of vinegar.
Above observation shows good quality wines have low volatile acidity which can infer it will taste well and smells good.
































6.3 Multivariate Analysis



Let's see how quality is varied by multiple variables

















6.3.1 Multivariate Pair Plots












In [137]:



    


sns.pairplot(df_wine,vars=['alcohol','sulphates','citric_acid','fixed_acidity','density'],hue='quality',diag_kind='kde')



    















    
Out[137]:








<seaborn.axisgrid.PairGrid at 0x1982cbd8748>










    




























Above Pair Plots show differnt relationships among differnt variable with overlaping quality.


    

  • Wine quality from 8-9 have lower fixed_acidity with high alcohol percentage
    • 

  • Most of the good quality wines have lower density with higher alcohol percentage
    • 

  • Lower sulphates with higher alcohol gives good quality wines
    • 



We will analyze further how other variables gives different quality variation  in following section

















6.3.2 Multivariate Scatter Plots












In [131]:



    


sns.scatterplot(x='fixed_acidity',y='volatile_acidity',hue='rating',data=df_wine,palette="husl")
plt.title('Quality based on Volatile_acidity and Fixed_acidity',fontsize=16)
sns.set(rc={'figure.figsize':(20,8.27)})



    















    




























Above observation illustrates that majority of the wines with average wine quality has fixed_acidity less than 10 with few execptions. Also most of the good and average wines have low volatile_acidity less than 0.8 however bad wines are having volatile acidity. 
Since the dataset is mixed with both white wines and red wines , we cannot concur that which bad quality wine has low fixed acidity or high volatile acidity. 
Based on research it can be assumed that white wines exihibits more fixed acidity to make them sparkle and give crispiness on the palate. 
Hence, we can say the wines falling in good category with high fixed acidity can be white wines and which has lower fixed acidity and volatile acidity can be assumed to good red wines













In [138]:



    


sns.scatterplot(x='alcohol',y='density',hue='rating',data=df_wine,palette="husl")
plt.title('Quality based on Density and Alcohol',fontsize=16)
sns.set(rc={'figure.figsize':(20,8.27)})



    















    




























We can observe from the above scatter plot that majority of the good wines have density lower than 1.00 and alcohol content between 12-14 %. On the other hand average wines have alcohol content between 8-12 %.
Also we can observe the how alcohol and density are highly negative corealted.
































7. Conclusion














The objective of this EDA was to analyze and understand what wine features may have most impact on the wine quality which was summed up to 3 categories viz bad,average and good. The dataset has majority of the values (76%) under 'average' wine quality, good wine quality was (20%) of the sample value whereas bad wine quality was around (4%) of the total sample value.
Although the dataset was clean with no missing values or negative values,however the dataset lacks the balance with quality interval.


Univariate analysis helped us to pick the features which are biggest influencer of quality and those were selected for further analysis. These features were :
    

  • Volatile Acidity
    • 

  • Alcohol
    • 

  • Sulphate
    • 

  • Citric Acid
    • 

  • Density 
    • 



Next phase of analysis revealed that :
    

  • Less volatile Acidity generally have high quality wines .
    • 

  • The lesser dense wine is will have more alcohol. 
    • 

  • Higher alcohol will have good quality of wine.
    • 

  • Citric acid around 0.5 or above will have average to good quality of wines.
    • 



Multivariate Analysis Observations:
    

  • Lower suplphate values with higher alcohol values considered to good quality wines
    • 

  • Most of the average to good wine quality falls under lower Fixed_acidity (<10) with lower (<volatile_acidity).
    • 

  • Lower density with higher alcohol percentage also derives to good quality wines. 
    • 



The dataset has few limitations as well like , it lacks the information of the category of wine like red or white. Due to mixed sample it was difficult to concur on quality based on features as both white and red wine will have different variations due to their base ingredients reaction. For eg. white wine may have slighter high fixed_acidity to give it a more crispiness in taste and also have high alcohol contents , that is why we see quite few exceptions in our data analysis.

















8. Actionable Insights














    

  • Medium quality wines have alcohol level between 9-12 and volatile_acidity level between 0.2 - 0.8 
    • 

  • High quality wines have alcohol level between 10-14 and volatile_acidity level between 0-2 -0.6. 
    • 

  • Low quality wines have alcohol level ranging between 9-12 and volatile_acidity level between 0.4 - 1.2 
    • 



We should add more variables like wine category, weather conditions, wine making process specifics to derive more valuable business insights.